Scholarlink

In order to achieve accurate prediction of short-term spatial and temporal distribution of crowd density in each area after the crowd enters the station space, and to solve the problems of risk warning and operational safety in railway station crowd management. This paper proposes a local density prediction model for pedestrians in public space areas based on graph embedding. By applying graph structure data, train schedule data and density distribution time series data. Pedestrian density prediction in spatial areas is achieved by attention Bi-directional LSTM model. The results show that the proposed method outperforms traditional prediction models such as ARIMA, RNN, LSTM and their ablation models. The RMSE error and MAE error are reduced from 12.048 to 1.639 and 11.592 to 1.148, respectively, which can be used as a reference for optimising the spatial layout of functional areas according to the pedestrian distribution and improving the spatial risk prevention ability in future railway station.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Graph Embedding and Attention Bi-LSTM Based Model on Prediction of Local Density Distribution of Crowd in Railway Station

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yongcheng Wang,
  • Dewei Li,
  • Zhicheng Dai,
  • Hong He

摘要

In order to achieve accurate prediction of short-term spatial and temporal distribution of crowd density in each area after the crowd enters the station space, and to solve the problems of risk warning and operational safety in railway station crowd management. This paper proposes a local density prediction model for pedestrians in public space areas based on graph embedding. By applying graph structure data, train schedule data and density distribution time series data. Pedestrian density prediction in spatial areas is achieved by attention Bi-directional LSTM model. The results show that the proposed method outperforms traditional prediction models such as ARIMA, RNN, LSTM and their ablation models. The RMSE error and MAE error are reduced from 12.048 to 1.639 and 11.592 to 1.148, respectively, which can be used as a reference for optimising the spatial layout of functional areas according to the pedestrian distribution and improving the spatial risk prevention ability in future railway station.